Run semantra server without parsing document again

This is a great project. Is there any way we can just parse the document once. Next time, when we need to search we just call something like `semantra server` to open the UI server and use the previous processed data?

1. Parse document and store in `semantra-dir`
2. run `semantra server`